Model-based algorithms for phenotyping from 3D imaging of dense wheat crops

High-throughput phenotyping requires the automation of the process of extracting relevant quantitative information from crop images. We achieve this goal for dense wheat crops by applying model-based data analysis techniques (including clustering and data-fitting) to high-resolution 3D point clouds obtained from structured light laser scanners. By performing experiments comparing our estimates of crop height, spike height and spike width with manual measurements, we demonstrate that our approach is promising even in the challenging context of dense vegetation.